Observed arrival · 2026-08-30
oneproof wants AI agents to leave evidence, not just logs
An open suite of tools for authorizing agent actions, detecting system changes, and tracing retrieved answers back to their source material.
Field notes
The project divides its evidence model into three stages: policy authorization, change monitoring, and retrieval provenance. Its visible examples include a ratified policy hash, explicit allow/refuse verdicts, a 74/74 baseline check, and a retrieval receipt naming chunk 41, section 4.1, and characters 1,204–1,451. onedoor is identified as the currently shipped component at version 0.6.2; the other two components are described with public or forthcoming verification work.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue